在图像配准和图像识别领域中, SIFT算法得到广泛应用。大量的噪声或是复杂的图像往往会导致匹配点的误判, 从而降低了图像匹配效率。提出一个新的比较邻域, 建立了一个新的“十字形”描述符并运用主成分分析法降低描述符计算时间, 利用双向匹配方式来减少伪匹配点, 以节省匹配时间。实验结果表明, 改进SIFT算法能够成功匹配图像, 应用于大型工业刀具的测量, 为大型物体测量提供了一种新的方法。

SIFT algorithm has become one of the most active research topics in computer vision in recent years. However, the existence of noise and complicated images also limit the range of using SIFT algorithm. A new neighborhood space to compare is proposed and a new descriptor incorporates a “crisscross”-like is established. The algorithm reduces dimensions to save time by using tow-way matching. This improvement of SIFT algorithm select out a large number of true matching points and increase the efficiency of matching. Experimental results show the improved SIFT algorithm can successfully match images. At the same time, the improved SIFT algorithm not only is applied to the measurement of large industrial tools but also provides a new method for the measurement of large objects.